INTRODUCTION

Receptor tyrosine kinases are widely expressed in the developing nervous system, where they play important roles in development of neurons and glia (1). Upon binding of membrane-bound or diffusable ligands to their extracellular domains, these enzymes autophosphorylate on cytoplasmic tyrosine residues, which serve as docking sites for src homology 2 (SH2)¹ domain-containing signal transduction proteins. In the developing fly eye, the receptor tyrosine kinase encoded by the sevenless gene is activated by a neighboring cell surface protein encoded by the boss gene (2), causing precursor cells to differentiate into photoreceptors (3, 4). In vertebrates, activation of receptor tyrosine kinases of the trk family by nerve growth factor, brain-derived neurotrophic factor, or neurotrophin-3 and neurotrophin-4/5 induces differentiation and survival of different neuronal populations (5). In the peripheral nervous system, glial growth factor/heregulin causes multipotent neural crest progenitors to differentiate into glia rather than neurons (6) by activating the erbB2/c-neu/HER2 receptor tyrosine kinase in the presence of the erbB4/HER4 tyrosine kinase (7).

Receptor tyrosine kinases were initially identified as homologs of retroviral oncogene products (8); thus, it is not surprising that mutations in proto-oncogenes that result in constitutive activation of normal receptor tyrosine kinases render these proteins oncogenic. Such activating mutations can occur within the coding region of the extracellular domain, for example in the retroviral oncogene v-erbB (8) and trk/nerve growth factor receptor gene (9). Alternatively, the mutation can be an amino acid substitution in the transmembrane region, as shown for the erbB/c-neu/HER2 gene in chemically induced rat glioblastomas (10). Another means by which a receptor tyrosine kinase can become constitutively activated is by overexpression, which may result in forced receptor dimerization within the plasma membrane (11). This is one mechanism that may contribute to the deregulation of growth in glial cell tumors. For example, the erbB2/c-neu/HER2 gene is overexpressed in certain human glioblastomas (12), while the epidermal growth factor receptor gene, c-erbB, is amplified to various degrees in human glioblastomas with the highest levels of expression correlating with poor prognosis (13).

Protein tyrosine phosphorylation in the developing neural retina is unusually active in Müller glia (14, 15, 16, 17). The Müller glial cell is the principal glial cell type in the vertebrate eye, where its chief function is to buffer the microenvironment following neuronal firing (18). The normally quiescent Müller cell proliferates in an unregulated manner in several pathological situations, including retinal detachment, diabetic retinopathy, proliferative vitreoretinopathy, and macular pucker (19). At least one receptor tyrosine kinase, the basic fibroblast growth factor receptor, is up-regulated in Müller glia that have been induced to proliferate in animal models of injury (20). Müller glia and all types of retinal neurons differentiate from a common progenitor by a poorly understood process that depends on growth factors and other local environmental cues such as cell-cell contact (21, 22, 23).

Because receptor tyrosine kinases are important in growth factor and cell contact recognition, it is likely that they regulate retinal cell differentiation. Protein tyrosine phosphorylation in the developing chick retina increases strikingly during the period of differentiation of retinal neurons and Müller glia and is most abundant in regions where Müller glial processes contact their neighbors (16, 17, 24). Immunoelectron microscopy of the outer chick retina with phosphotyrosine antibodies revealed that phosphotyrosine-modified proteins accumulate predominantly in Müller glia and that they are located in the Müller glial plasma membrane at sites of contact with adjacent Müller glial processes and photoreceptors (24).

To identify protein-tyrosine kinases that might be responsible for the elevated protein tyrosine phosphorylation in developing Müller cells, we used reverse transcriptase-polymerase chain reaction (PCR) with primers specific to highly conserved regions shared within the catalytic domain of receptor and nonreceptor-class protein-tyrosine kinases to amplify partial cDNAs encoding tyrosine kinases expressed in Müller glia-enriched cultures from embryonic chick retina. Such an approach has been used with success to identify novel tyrosine kinase genes expressed in nonneural as well as neural cells (25, 26).

Here we describe the discovery of a novel receptor-type tyrosine kinase, termed Rek (retina-expressed kinase), a new member of the Axl/Tyro3 family of receptor tyrosine kinases. The Axl/Tyro 3 family includes receptor tyrosine kinases encoded by axl (ufo, ark) (27, 28, 29, 73, 85), tyro3 (sky, brt, rse, tif) (26, 30, 31, 32, 33), c-eyk (34), and c-mer (35). There is evidence that these kinases have transforming potential. Axl was originally identified as a protein encoded by a transforming gene from primary human myeloid leukemia cells (27). Axl is overexpressed in a number of different tumor cell types and transforms mouse NIH3T3 fibroblasts (27). Experimental overexpression of Tyro3 causes anchorage-independent growth of Rat-2 fibroblasts (36). Also, the murine homolog of sky has been shown to be expressed at elevated levels in mouse mammary tumors (37). c-eyk is a chicken proto-oncogene that was first identified as the retroviral transforming gene v-ryk (34). A close relative, c-mer, is a human proto-oncogene expressed in malignant B- and T-lymphocytic cell lines (35).

The hallmark of the Axl/Tyro3 family is an extracellular region consisting of two immunoglobulin-like (Ig) and two fibronectin III (FN) domains. These domains are found in cell recognition molecules such as the neural cell adhesion molecules L1 and NCAM (38) and certain receptor tyrosine phosphatases (39, 40). Homotypic and heterotypic binding involving extracellular determinants have been demonstrated for some Axl/Tyro3 family members. Homotypic binding has been shown to activate the Axl tyrosine kinase (41), an event that may be important in signaling cell adhesion. Axl (42) and, to a lesser extent, Tyro3 (43) are also activated by a heterophilic ligand, Gas6, a vitamin K-dependent protein that is up-regulated during growth arrest in confluent fibroblast cultures. Protein S, which bears significant homology to Gas6, was found to be a heterophilic ligand for Tyro3 (44). Protein S is an anticoagulant in serum and a mitogen for smooth muscle cells (45), but its up-regulation in Schwann cells following nerve injury suggests that it may also serve as a neural growth or differentiation factor (44).

The molecular cloning, structural analysis, and expression of the Rek receptor tyrosine kinase in developing chick neural tissues is described here. It is also demonstrated that overexpression and activation of Rek in mouse NIH3T3 fibroblasts leads to cell transformation, indicating that the rek gene has oncogenic potential that might contribute to malignant growth of nervous system tumors.

MATERIALS AND METHODS

Primary cultures enriched for Müller glia were prepared from embryonic day 10 chicken retinas as described previously (17, 46). Retinas were dissected free of sclera-choroid-pigmented epithelium and cells seeded on tissue culture dishes in DMEM, 10% fetal calf serum. By 7-10 days, a mixed culture developed consisting of a monolayer of flat cells underlying a network of neuronal fascicles and cell clumps, which were removed by gentle mechanical agitation. As described previously, the identification of the flat cells as Müller glia is supported by abundant intermediate filaments (47), staining with vimentin antibodies (48), lack of binding to tetanus toxin, [³H]thymidine incorporation, GABA and glutamate uptake (49, 50), expression of a filamin-related protein (51), and the presence of carbonic anhydrase and glutamine synthetase (52). The glial-like cells do not show induction of glutamine synthetase by hydrocortisone, a response characteristic of Müller cells in vivo (52), and are negative for expression of glial fibrillary acidic protein, perhaps due to the absence of necessary intercellular interactions in the monolayers. They will be referred to as Müller glia-like cultures, but they may not represent fully differentiated Müller cells or their precursors.

Hatchling chicks (White Leghorn, NC State) were anesthetized with Ketamine and sacrificed by decapitation. Staging of chick embryos was done according to the guidelines of Hamilton and Hamburger (53). For RNA, tissues were rapidly excised, frozen in liquid nitrogen, and stored at 80 °C until use. Poly(A)⁺-selected RNA was isolated using the FastTrack RNA isolation kit (Invitrogen).

Poly(A)-selected RNA (1 µg) from Müller glial cell cultures was reverse transcribed using random hexamers as described by O'Bryan et al. (27) in 10 mM Tris, pH 8.3, 50 mM KCl, 1.5 mM MgCl₂, 0.001% gelatin, 625 µM each deoxynucleoside triphosphate, 20 units of RNAsin (Promega), 10 mM of dithiothreitol, and 200 units of Moloney murine leukemia virus reverse transcriptase (final volume, 20 µl). The reaction was incubated for 10 min at room temperature and then for 45 min at 45 °C. This first strand cDNA (1 µl) was amplified in a 100-µl reaction containing 10 mM Tris HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl₂, 200 µM each dNTP, 0.001% gelatin, 1 µg each primer, and 1.25 units of Taq polymerase (Promega). The degenerate oligonucleotide primer sequences of Wilks (25) were used corresponding to the following amino acid sequences: PTK1, CGG ATC CAC (A/C)GN GA(C/T) (C/T)T; PTK2, CT(G/A)CA(G/C) ACC AGG A(A/T)A CCT TAA GG. Reaction conditions were as follows: 1.5 min at 95 °C (denaturing), 2 min at 37 °C (annealing), and 3 min at 63 °C (elongation). The PCR products were ligated into the plasmid pBluescript (Stratagene) and transfected into DH5 Escherichia coli cells (Life Technologies, Inc.), and colonies were selected after induction with 5-bromo-4-chloro-3-indoyl -D-galactoside and isopropyl-1-thio--D-galactopyranoside.

To produce a DNA fragment of suitable length for a hybridization probe, a modification of the 3-RACE method (54) was used to generate a 1.4-kb fragment representing the 3-end of rek cDNA extending from the sequence encoding the IHRDL motif in the catalytic domain to the poly(A) tail. For this purpose, an oligo(dT)-primed ZAPII cDNA library was generated from Müller glia-enriched cultures using a cDNA cloning kit (Stratagene). A nondegenerate primer specific to rek (seal-1, ATG CTG GAT GAG AAC ATG AAT; corresponding to amino acids MLDENMN; residues 649-655) was used as the sense primer, and an oligonucleotide (pBlu5) specific to the pBluescript phagemid between the XhoI site and the T7 promoter (ATA GGG CGA ATT GGG TAC) was the antisense primer. PCR was carried out as above with 0.75 mM MgCl₂, using the following reaction conditions: 5 min at 94 °C (denaturing), 5 min at 60 °C (annealing), and 4 min at 72 °C (extending) for 1 cycle, followed by 45 s at 94 °C, 45 s at 60 °C, and 4 min at 72 °C for 35 cycles. The PCR product was directly ligated into the pCR plasmid (TA cloning vector, Stratagene), DH10 E. coli cells were transformed, and colonies were selected. The insert sequence from one of the 3-RACE clones was sequenced, verifying that it specified the Rek catalytic domain.

An oligo(dT)-primed cDNA library in  gt10 from chick embryonic brain (day 13) (Barbara Ranscht, Burnham Cancer Research Foundation) was screened at high stringency using the ³²P-radiolabeled 3-RACE clone (1.4 kb) as probe. Out of the 28 positive primary clones obtained, 15 clones survived plaque purification through secondary and tertiary screening. Subclones of the longest clone (approximately 4 kb) were generated by exonuclease III digestion, and DNA sequencing was completed for both strands. This clone contained a long open reading frame, a 3-untranslated region containing an internal A-rich sequence, and a poly(A) addition site, but it lacked a start codon and signal peptide. The additional 5-sequence was obtained by rescreening the brain library using as probe a PCR fragment amplified from the 5-region of the 4-kb clone. Twelve clones were plaque-purified through tertiary screening, and their 5-regions were sized by PCR. The clone with the longest 5-region was sequenced and found to contain the full rek coding sequence, including a putative start codon and signal peptide sequence. The remainder of the sequence was identical to the 4-kb cDNA clone. Interestingly, the clone specifying the entire protein sequence (3.3 kb) represented a cDNA that had been internally primed at an A-rich sequence in the 3-untranslated region of the mRNA.

Two methods of DNA sequence analysis were used. Manual DNA sequencing by the dideoxy chain termination method was carried out using Sequenase T7 polymerase (U.S. Biochemical Corp.) with T3 and T7 primers. Automated sequencing was performed in the UNC-Chapel Hill Automated DNA Sequencing Facility (Dr. Laura Livingstone, Director), which employs a model 373A DNA sequencer (Applied Biosystems). Both sense and antisense strands were sequenced. Sequences were compared by the FASTA program to the GenBank^TM/EMBL and SwissProt data bases using the GCG software package. Contiguous clones were aligned using the GAP program and then merged using the ASSEMBLE program. The PILEUP program was used to compare the Rek amino acid sequence with other members of the Axl/Tyro3 family. The phylogenetic tree was generated by the computer program Phylogenetic Analogy Using Parsimony (PAUP)(55).

RNA was separated on denaturing 1% agarose, 2% formaldehyde gels and subjected to blot hybridization on Hybond N membranes at high stringency (83). Hybridization using a ³²P-radiolabeled DNA probe generated by PCR amplification of the 1.4-kb 3-RACE clone was carried out at 42 °C overnight in 5 × SSC, 40% formamide, 5 × Denhardt's solution, 0.1% SDS, 1 mM NaH₂PO₄, and 200 µg/ml boiled salmon sperm DNA. Washes were as follows: 6 × SSC, 0.1% SDS for 30 min at 42 °C, 2 × SSC, 0.1% SDS for 30 min at 42 °C, and 1 × SSC, 0.1% SDS for 20 min at 55 °C. The filters were exposed to film at 80 °C for 6 days with intensifying screens. Normalization of the amount of RNA loaded was confirmed by hybridizing to an actin probe, which was generated by PCR amplification of Müller glial cDNA with actin specific primers as described in O'Bryan et al. (27).

Chicken genomic DNA was partially digested with EcoRI or HindIII, and fragments were separated by agarose gel electrophoresis and blotted to nitrocellulose. Filters were hybridized to two EcoRI/BamHI fragments (0.4 and 0.9 kb) encoding the entire extracellular and transmembrane domains (nucleotides 1-1602), which were ³²P-labeled by random priming. Filters were hybridized overnight in 50% (v/v) formamide, 4 × SSC, 5% (w/v) Denhardt's solution, 20 mM sodium phosphate, pH 7.5, at 42 °C and washed two times for 20 min each in 2 × SSC, 0.5% SDS at room temperature, once in 2 × SSC, 0.1% SDS for 20 min at room temperature, and once in 2 × SSC, 0.1% SDS for 30 min at 50 °C. Filters were exposed to x-ray film for 24 h.

Two different BamHI-HindIII fragments of the 3.3-kb rek cDNA clone were used to express Rek fusion proteins as antigens. One (for antibody A) encoded 305 amino acids of the kinase domain (residues 459-764), and the other (for antibody B) encoded the carboxyl-terminal 100 residues. The fragments were subcloned into the pLC24 vector (56), and fusion proteins with 98 amino acids of the MS2 polymerase protein were expressed in E. coli upon temperature shift from 28 to 42 °C. The proteins were purified by preparative SDS-PAGE and used to inoculate rabbits. Rek fusion proteins (150 µg) in complete Freund's adjuvant were used for the primary injection and for each of three boosts in Freund's incomplete adjuvant. Antisera were screened by Western blotting against the fusion proteins. Where indicated, antibodies were purified from IgG preparations by immunoaffinity on an Affi-Gel column (Bio-Rad) to which the fusion proteins were covalently coupled. Antibodies were eluted with 50 mM diethylamine and dialyzed against phosphate-buffered saline.

The entire 3.3-kb cDNA encoding the complete Rek protein was cloned into the EcoRI site of the vector pSG5 (57) and transfected into bacteria. Clones were selected with rek cDNA in either the sense or antisense orientation with respect to the SV40 promoter of pSG5. COS-7 cells (in 100-mm dishes) were transfected for 6 h in OptiMEM medium (Life Technologies, Inc.) with the sense or antisense construct (5 µg) using lipofectamine (36 µl; Life Technologies, Inc.). Cells were washed and incubated in fresh medium and then passaged 24 h later onto new 100-mm dishes. Metabolic labeling was carried out for 14 h with an ³⁵S-protein labeling solution (EXPRE³⁵S³⁵S, DuPont NEN) containing radiolabeled L-methionine and L-cysteine in DMEM (without methionine and cysteine), 10% dialyzed fetal bovine serum. Cells were lysed in RIPA buffer and immunoprecipitated with preimmune serum or rek antiserum (350 µl of lysate, 12 µl of serum) or Rek antiserum preadsorbed of Rek-specific antibodies by passing IgG through an Affi-Gel column to which the antigen-fusion protein was coupled. Proteins were separated by SDS-PAGE and visualized by fluorography on x-ray film (1-week exposure).

Rek was immunoprecipitated from either RIPA (24) or Brij 97 detergent extracts of transfected COS cells or Rek lysis buffer (RLB) extracts of transfected NIH3T3 cells using antibodies prepared as described above and protein A-Sepharose. The composition of Brij 87 buffer was 1% Brij-97, 10 mM Tris, pH 7.4, 150 mM NaCl, 1 mM NaEDTA, 1 mM NaEGTA, 10 mM NaF, 200 µM Na₃VO₄, 500 µg/ml Pefabloc, 0.01% leupeptin, 0.11 trypsin inhibitory units/ml aprotinin. The composition of RLB was 50 mM HEPES, pH 6.5, 150 mM NaCl, 1.5 mM MgCl₂, 10% glycerol, 1% Triton X-100, 1 mM Na-EGTA, 200 µM sodium orthovanadate, 10 mM NaF, 0.11 trypsin inhibitory units/ml aprotinin, 0.01% leupeptin. Immune complexes were washed in RIPA, Brij 96, or RLB, respectively, and incubated with a reaction buffer containing 50 nM to 1 µM [-³²P]ATP-Mn²⁺ in a 30-µl volume as described for 30 min at 37 °C (15). Products were separated by SDS-PAGE, and gels were fixed in 10% acetic acid, 30% methanol. Phosphoserine and phosphothreonine residues were selectively dephosphorylated by treating gels with 1 M KOH at 55 °C for 45 min and then fixing again for 10 min. Gels were dried and set up for autoradiography at 70 °C using x-ray film. In some experiments, Rek was subjected to immunoblotting with phosphotyrosine antibodies (Upstate Biotechnology, Inc.; Transduction Laboratories, Inc.) and detection by enhanced chemiluminescence (Amersham Corp.).

For Western blotting, Rek was immunoprecipitated from RLB extracts (500 µg; 1 ml) of chick embryonic retina or 50 µg of NIH3T3 cells using 20 µl of Rek antiserum B and protein A-Sepharose. After SDS-PAGE, proteins were transferred to Immobilon filters, and the filters were blocked with 3% fish gelatin or 3% milk protein in 120 mM NaCl, 50 mM Tris-HCl (pH 7.4), 0.1% Tween 20. Western blotting was carried out using Rek antiserum B (1:500) with detection by enhanced chemiluminescence.

NIH3T3 cultures were maintained in DMEM with 10% fetal calf serum. The rek cDNA was subcloned into the pLXSN retroviral expression vector. The axl expression vector was constructed in pLXSN as described previously (27). Five µg of each plasmid or pLXSN alone were transfected into an NIH3T3 cell line as described (27). Briefly, 2 × 10⁵ cells were plated in each 60-mm tissue culture plate on the day prior to transfection. The next day, DNAs were precipitated in the presence of calf thymus DNA (20 µg) using the calcium phosphate method and then added to cells. The cells were incubated at 37 °C for 5-8 h and then glycerol-shocked to increase the efficiency of DNA uptake. After 2 days, the cells were passaged and grown in the antibiotic G418 (600 µg/ml final concentration), and drug-resistant colonies were selected after 2 weeks. G418-resistant colonies were combined, passaged into 60-mm plates, and allowed to reach confluence. The appearance of morphologically transformed foci were visualized by phase contrast microscopy after 3-4 weeks.

RESULTS

We used reverse transcriptase-PCR with primers flanking a highly conserved region within the tyrosine kinase catalytic domain (25) to identify genes that coded for tyrosine kinases whose RNA transcripts were present in primary cultures enriched in Müller glia-like cells from embryonic chick retina. The insert sequences of 80 PCR clones obtained (each 210 base pairs in length) were subjected to DNA sequence analysis. This resulted in the identification of 11 different sequences encoding tyrosine kinases expressed in the retinal cell cultures, eight of which were identical to previously described tyrosine kinases. Five of these represented sequences encoding receptor class tyrosine kinases: elk, fibroblast growth factor 2-receptor (cek3, bek), hepatocyte growth factor receptor (met), -platelet-derived growth factor receptor, and insulin-like growth factor-I receptor. Three represented nonreceptor tyrosine kinases: fyn, CSK (carboxyl-terminal src kinase), and c-abl. The expression of the fyn gene in Müller glia-enriched cultures was previously demonstrated by immunoblotting (24). The predicted amino acid sequence of each clone was at least 94% identical to that of the designated protein-tyrosine kinase in the data base, identifying it as a putative chicken homolog. One of the PCR clones, appearing twice in the screen, encoded the catalytic region of a potentially novel receptor-like tyrosine kinase (Rek).

To obtain full-length rek cDNAs, a chick embryonic brain (day 13)  gt10 cDNA library was screened as described under "Materials and Methods." A composite sequence of cDNA comprising the full coding sequence of the rek protein and the 5- and 3-untranslated regions is shown in Fig. 1. The cDNA sequence (4961 nucleotides) comprised a GC-rich 5-untranslated region (228 nucleotides), a coding sequence (2619 nucleotides), and a 3-untranslated region (2116 nucleotides). It was not determined whether the complete 5-untranslated sequence was represented in the clones obtained. The 3-untranslated region included an internal poly(A)-rich sequence (nucleotides 3498-3513), an AATAAAA polyadenylation recognition site, and two ATTTA sites, which have been associated with mRNA instability in cytokine and growth factor transcripts (57).

The rek cDNA sequence predicts a precursor protein of 873 amino acids (molecular weight 96,370) with a putative signal peptide of 28 residues, an extracellular region of 385 residues, a hydrophobic transmembrane domain of 25 residues, and an intracellular region of 435 residues. The cytoplasmic region comprises a catalytic domain conserved among protein-tyrosine kinases followed by a divergent carboxyl-terminal sequence. The predicted amino acid sequence contains 2 in-frame methionines that are candidate initiating amino acids based on Kozak consensus rules (59). The first methionine marks the beginning of the characteristic signal peptide sequence. The signal peptide has a positively charged amino-terminal region, followed by a hydrophobic region rich in leucines and alanines, and a polar carboxyl region satisfying the (3, 1) rule for defining the signal peptidase cleavage site (60). Supporting the assignment of the initiating amino acid as the first methionine, there is an in-frame termination codon (TAG) lying in a highly GC-rich region located 168 base pairs 5 of the initiating codon.

The amino acid sequence of the extracellular region revealed two tandem Ig domains (37) followed by two tandem FN III motifs (61). The first Ig domain is a typical C2 type structure (37) with a conserved cysteine (amino acid 52) followed by a tryptophan (amino acid 64) and a cysteine (amino acid 105) within the consensus motif DXGXYXC. The second Ig domain also has a conserved sequence of a cysteine (amino acid 148), two tryptophans (residues 161 and 162), and a cysteine (residue 191), but the latter cysteine is not located within the characteristic consensus motif. The two FN III repeats share residues conserved in the fibronectin molecule and in other proteins with FN III repeats (62). There are eight potential N-linked glycosylation sites (NX(S/T)) dispersed among the Ig and FN III domains, suggesting that the mature protein is highly glycosylated (63).

Following the extracellular domain is a hydrophobic transmembrane segment (amino acids 414-438) (64). The tyrosine kinase domain (amino acids 505-672) exhibits signature motifs conserved among other protein kinase catalytic domains, including GXGXXG, which binds ATP-Mn²⁺, and the triplet motifs RDL, DFG, and ALE, which function in the catalytic loop (65). Diagnostic of a receptor-like kinase is the presence of PVKWLALE and AARN sequences carboxyl-terminal to the RDL sequence. There are two potential tyrosine autophosphorylation sites adjacent to each other in the catalytic domain and a third tyrosine located 4 positions amino-terminal to these residues. These three tyrosine autophosphorylation sites are found in all members of the insulin receptor family (66). The catalytic domain contains a kinase insert (IGENPFN) typical of the insulin receptor superfamily (amino acids 607-613).

The catalytic and carboxyl-terminal regions possessed tyrosine residues within two consensus motifs that could serve as binding sites for SH2 domain-containing signaling proteins. The first site (YDLM; residues 749-752) was a potential binding motif that has been shown in synthetic phosphopeptide libraries to prefer sequences in the SH2 domain of the p85 regulatory subunit of phosphatidylinositol-3-kinase (67). The second site (YVNI; residues 791-794) was a potential binding motif for the SH2 domains of Grb2/sem5 (67). Three additional tyrosine residues (residues 832, 834, and 863) were located carboxyl-terminal to the second putative SH2 domain binding site, and may also function in signaling.

Comparison of the rek sequence with the GenBank^TM data base revealed that Rek was most closely related to Tyro3, with 61% identity in nucleotide sequence to mouse Tyro3 (36). Rek displayed 66% identity in amino acid sequence to mouse Tyro3 over the entire protein sequence, 56% identity in the extracellular region, and 87% identity in the catalytic domain but only 36% identity in the carboxyl-terminal region. Rek exhibited lower overall homologies in amino acid sequence with human Axl (43% identity; Ref. 27), chicken Eyk (41%; Ref. 34), and human Mer (41%; Ref. 35). Alignment of the Axl/Tyro3 family members illustrated the highest degree of conservation in the catalytic domain and second immunoglobulin-like domain but little conservation in the carboxyl-terminal tail (Fig. 2). An exception was the second putative SH2 binding site and an arginine-tyrosine sequence that was conserved in the cytoplasmic domain in all family members.

The entire amino acid sequences of members of the Axl/Tyro3 family were compared phylogenetically using the computer program PAUP (55) (Fig. 3). This statistical analysis, which reflected sequence relatedness, placed Rek in the Axl/Tyro3 family but distinguished Rek and Tyro3 as distinct gene products. By this analysis, Rek is more closely related to Tyro3 than to Axl and is more distantly related to Eyk and Mer.

To demonstrate that the 3.3-kb rek cDNA encoded a functional protein, the complete rek cDNA clone (an EcoRI fragment) was subcloned into the eukaryotic expression vector pSG5, and the construct was transfected into COS-7 cells for transient expression. Transfected COS cells were metabolically labeled with [³⁵S]methionine and [³⁵S]cysteine, lysed in RIPA buffer, and immunoprecipitated with Rek antibodies. Rek antibodies immunoprecipitated a single ³⁵S-radiolabeled protein of 106 kDa from COS cells that were transfected with sense but not antisense rek plasmids (Fig. 4A). Neither preimmune serum (lanes 1 and 4) nor Rek antiserum preadsorbed with purified Rek fusion protein (lanes 3 and 6) immunoprecipitated the 106-kDa protein. Moreover, the 106-kDa protein was not immunoprecipitated from ³⁵S-labeled nontransfected COS cells (lane 7). The size of this protein (106 kDa) was in good agreement with the predicted size of an unprocessed Rek precursor protein (96 kDa). The 106-kDa protein was immunoprecipitated with Rek antibodies from RIPA extracts of Rek-expressing COS cells and subjected to immune complex protein kinase assays with [-³²P]ATP (Fig. 4B). In this assay, the 106-kDa protein was the principal protein autophosphorylated. This phosphorylation may have represented a basal level of ligand-independent autophosphorylation or constitutive activation due to forced dimerization of the overexpressed receptor (11).

To examine the tissue-specific pattern of expression of rek, Northern blot analysis was carried out with poly(A)⁺ RNA isolated from different tissues of embryonic and hatchling chicks. The size of the composite cDNA sequence of rek excluding the poly(A) tail (4.9 kb) approximated the size of the rek transcript (5.5 kb) on Northern blots. A single 5.5-kb message was present in hatchling chick brain, retina, and kidney but was not detected in spleen, heart, liver, or skeletal muscle (Fig. 5A). Comparison to actin mRNA indicated that the hatchling brain and retina had approximately the same level of rek transcripts, while kidney had half this amount. The similar levels of expression in brain and retina suggested that rek was not restricted to a small subpopulation of neurons or glia. rek transcripts in kidney could reflect expression in adrenal medullary endocrine cells, which have the ability to transdifferentiate to sympathetic neurons (68, 69). A rek transcript of the same size was observed at embryonic day 13 in brain and retina but not heart, liver, or skeletal muscle (not shown). rek transcripts in the developing chick neural retina were seen at embryonic day 8 (stage 34), day 10 (stage 36), and hatching (Fig. 5A). During this time, neural progenitors differentiate into Müller glia and all types of retinal neurons (71, 72). Normalization to actin mRNA indicated that the relative levels of rek transcripts were approximately the same from E8 to hatching. Poly(A)⁺ RNA isolated from Müller glia-enriched cultures contained low but detectable levels of rek transcripts (Fig. 5A).

Rek protein was expressed in the developing chick retina at embryonic days 6-13 as a principle band of 140 kDa with a minor band of 120 kDa (Fig. 5B). Differences in the extent of N-glycosylation could account for the differences in molecular weight of the Rek protein in retina and COS cells. Similar results are seen with Axl when expressed in insect cells versus fibroblast cells (27). The relative levels of rek protein decreased at embryonic days 11 and 13, suggesting that developmental regulation may occur posttranscriptionally. Longer exposure of the film showed that Rek protein was still present at these stages and at embryonic day 15, although levels were lower than at embryonic days 9 and 10. This pattern of expression was consistent with a role for Rek in differentiation or proliferation of retinal neural cells.

Southern blotting with a rek cDNA probe corresponding to the extracellular region and transmembrane domain showed a single hybridizing band in chicken genomic DNA digested with EcoRI and two bands in chicken DNA digested with HindIII under relatively high stringency (Fig. 6). This indicates that the rek gene is a single copy gene and that there are not likely to be closely related genes in the chicken with substantial identities in the putative ligand binding region. Under the same conditions, no hybridization to mouse, rat, or human genomic DNA was observed.

In view of the demonstrated transforming activity of Axl (27, 72) and Tyro3 (36), the ability of Rek to transform fibroblasts in culture was assayed. NIH3T3 cells were transfected with an expression vector (pLXSN) encoding the Rek or Axl receptor to assess the transforming potential of Rek relative to Axl. As shown in Fig. 7, overexpression of Rek or Axl caused morphologic transformation of NIH3T3 cells. Cells transfected with the axl expression plasmid showed altered morphology. However, the phenotype of the cells was qualitatively different from that of the rek-transformed cells (Fig. 7). rek-transformed cells were only slightly refractile but with an obvious spindle-like appearance, whereas axl-transformed cells were more refractile and rounded. Additionally, rek-transformed cells appeared in the cultures with a latency of 3-4 weeks, 1-2 weeks later than in axl-transformed cultures. However, transformed foci were abundant in each dish of rek-transfected cells. Similar results were observed in 13 independent transfections. These findings suggested that rek, like axl, is capable of inducing morphological transformation of NIH3T3 cells. However, rek transforming activity appears to be less potent compared with that of axl.

Immune complex kinase assays with Rek antibodies showed that high levels of Rek were expressed in the rek-transformed NIH3T3 cell cultures (Fig. 8A). Cultures resulting from two independent rek transfections showed different levels of overexpressed Rek kinase. In vitro phosphorylated Rek appeared as it did in the chick retina, as a principal band of 140 kDa and a secondary band of 120 kDa. Similarly, Axl is expressed in transfected cells as a mature 140-kDa protein that can be converted to a partially glycosylated 120-kDa protein and an unglycosylated protein of 104-kDa by N-glycanase treatment (84). A small amount of kinase activity immunoprecipitating with Rek antibodies was seen in cells transfected with vector alone and axl-transformed cells. This result may indicate that NIH3T3 cells express low levels of endogenous Rek or a cross-reacting kinase.

The in vivo state of Rek phosphorylation at tyrosine in the transformed NIH3T3 cells was evaluated by immunoprecipitating Rek from two different rek-transformed cultures and immunoblotting with phosphotyrosine antibodies (Fig. 8B). Both rek-transformed cultures contained elevated levels of phosphotyrosine-modified Rek protein, which migrated as a predominant 140-kDa protein. The 120-kDa Rek protein detected in immune complex assays was present on longer exposure of the phosphotyrosine blots. Cells transfected with vector alone and axl-transformed cells did not contain detectable levels of phosphotyrosine-modified Rek protein. These findings show that overexpression of rek results in constitutive kinase activation in NIH3T3 cells and that kinase activation correlates with cell transformation.

DISCUSSION

Rek, a putative novel member of the Axl/Tyro3 family of receptor tyrosine kinases, was identified by molecular cloning and analysis of cDNAs isolated from an embryonic chick brain library. Sequence homology and distinguishing structural features of the predicted Rek protein placed it in the insulin receptor superfamily, which includes receptor tyrosine kinases with neural developmental functions such as the Drosophila sevenless protein (3, 4) and the trkA/nerve growth factor receptor (5). rek expression was elevated in developing neural tissues and corresponded to the period of differentiation of Müller glia and retinal neurons, suggesting a function in neural differentiation.

Although both Rek and Tyro3 tyrosine kinases are expressed in neural tissues, several lines of evidence indicate that they are probably the products of different genes. The chicken Rek tyrosine kinase displayed a 66% identity in predicted amino acid sequence to mouse Tyro3 over the entire protein sequence, which included the contribution of the highly conserved catalytic region. Significantly lower identities were observed in the extracellular (56%) and carboxyl-terminal (36%) regions. Comparison of chicken Rek with the human Tyro3 ortholog, Rse (33), revealed a degree of overall identity (68%) that was substantially less than that between the chicken and human EGF receptors (88%) (35). Within the extracellular region, the amino acid identity of Rek and Rse was also lower (61%) than that of the chicken and human EGF receptors (84%). Moreover, this level of identity is also found in the extracellular region of the chicken EGF receptor and human ErbB2/c-Neu/HER2 receptor tyrosine kinase (63%), a closely related but distinct EGF receptor family member (10). Finally, there is striking divergence in the carboxyl-terminal regions of Rek and Rse despite a conserved cysteine at the carboxyl terminus (33%). Notably, Rek is expressed at high levels in the retina, a site of low Tyro3 expression (43). However, since Rek is substantially more homologous to Tyro3/Rse than to other subfamily members, it is either the Tyro3/Rse ortholog or a second member of a Tyro3/Rse subfamily. Until the mouse ortholog of rek is cloned or another chicken gene is identified that is more closely related to tyro3/rse than is rek, the possibility remains that rek could be the chicken ortholog of tyro3/rse.

The Ig and FN III domains of Rek, like those of neural cell adhesion molecules, are likely to bind different molecular components. The different Ig and FN III domains of L1 (74) and F3/F11/contactin (75) have distinct functions in neural cell adhesion, neurite outgrowth, and neurite repulsion and bind different different homophilic and heterophilic partners. Sequence differences in the extracellular region of Rek, particularly in the first Ig and both FN III domains, may indicate that these domains have specificities different from those of Tyro3 or Axl. At this time, little is known about the homo- or heterotypic ligand binding properties of Rek, although preliminary experiments have not revealed an ability of Gas6 to activate Rek kinase. The second Ig domain of Rek was more highly conserved than other extracellular domains, which could indicate that there are some molecular interactions that this family has in common.

The existence of two consensus SH2 binding motifs in the Rek carboxyl-terminal region predicts an ability to activate several intracellular signaling pathways. The first motif (YDLM) is predicted to bind phosphatidylinositol-3-kinase, whereas the second motif (YVNI) is predicted to bind Grb2/Sem5, which activates Ras signaling pathways, although other SH2 domain-containing proteins are not ruled out (67). The final tyrosine in Rek is highly conserved in this family and could be an additional phosphorylation site. Although the tyrosine residues at all three sites are conserved in Rek, Tyro3, Axl, and Eyk, residues following the tyrosine at positions 1-3 are not highly conserved, and these residues are known to be determinants of binding specificity for signal transduction proteins (67). It has been shown that the second SH2 binding site in Axl (YVNM) binds both phosphatidylinositol-3-kinase and Grb2/Sem5 in cells (77), whereas the lack of a methionine at position 3 in the second putative binding site in rek (YVNI) should result in a much lower selectivity for phosphatidylinositol-3-kinase (67). The 4-6-positions also influence specificity, and nonconservative amino acid substitutions in Rek and Tyro3 are found in these positions at both putative SH2 domain binding sites. Thus, Rek may activate a distinct but overlapping set of signaling pathways compared with tyro3 and other family members. An interesting possibility is that Rek and Tyro3 may form heterodimers within the plasma membrane of neural cells. Ligand-induced heterodimerization has been demonstrated in the EGF and PDGF receptor families, where it can recruit different combinations of SH2-containing proteins for intracellular signaling. Such a mechanism provides a means for increasing the range of cellular responses from a limited number of ligands and receptor kinases (Refs. 11, 75, 76 and references therein).

The restricted tissue expression of rek suggests a possible role in neural cell growth or differentiation. During development of the neural retina, rek-specific transcript and protein levels were elevated throughout the period during which Müller glia and retinal neurons differentiate from a common proliferative progenitor (70). rek expression has not yet been studied at the cellular level, but its presence in Müller glia-enriched cultures suggests that it might be a product of differentiating Müller glia or neural progenitors in vivo. It is interesting to speculate that activation of Rek kinase by homo- or heterotypic binding between apposing Müller cell surfaces might be responsible for the increased protein tyrosine phosphorylation observed at sites of cell-cell apposition in Müller glial processes in the chicken retina (24). rek expression in the kidney could be related to its presumed function in brain and retina if Rek protein were found to be localized to sympathoadrenal precursors rather than to renal cells (68, 69).

Overexpression of rek in mouse NIH3T3 fibroblasts resulted in cell transformation and kinase activation. This appears to be an important feature of the axl/tyro3 family of receptor tyrosine kinases, since both Axl (27) and Tyro3 (36) are transforming when overexpressed. Transforming ability has also been demonstrated for an overexpressed eph receptor tyrosine kinase (79), which functions in retinal axon guidance (80, 81). The degree of transformation induced by overexpression of rek or axl was similar to that of overexpressed EGF receptor in NIH3T3 cells.² It is important to note that, since a ligand for Rek has not yet been identified, it is possible that Rek is activated by a molecule present in growth medium containing fetal calf serum, especially in light of the reported activation of Tyro3 by the serum protein, protein S (43). However, Sky, the human homolog of Tyro3, has recently been shown to be activated in a ligand-independent manner (85), suggesting that Rek activation may also be ligand-independent. Other receptor tyrosine kinases such as ErbB2/Neu/HER2 cause potent ligand-independent cell transformation (78). Because exogenous EGF significantly increases the transforming ability of the EGF receptor (82), it is possible that putative ligand-independent rek transforming ability would be potentiated in the presence of an activating ligand. The ability of rek to transform NIH3T3 fibroblasts when overexpressed demonstrates that this receptor tyrosine kinase has oncogenic potential. In this regard, it is interesting to consider that mutations giving rise to rek overexpression or activation in vivo might contribute to the growth of malignant nervous system tumors.